Using WS2812B strips as a matrix

[GrayOrion]

Hi!
I have a bunch of 4' long WS2812B strips mounted on strips of wood that are mounted with 3d printed clips, allowing me to use them in different configurations. I would love to use them as a matrix as well (to create wider effects or to display text). For the life of me I cannot figure out if there is an actual simple way to define a matrix zigzagged led strips in the projects (I have an din and dout on the ends, so I would just ideally chain them and inject some power ever so often).
Help please!
Oh, I am running them on Seeed Studio esp32-s3 which was not too hard to do with this project, but have not added too many effects so far...

[robertlipe]

Hi, and welcome. The literal answer to the question you asked is "no". But that's probably not the answer you're looking for. The pixel-blasting code likes to think that there's one giant pixel array that gets sent to a device. You've surely seen references to gfx->leds[] everywhere in the various effects. The "trick" is to slice that strip up, exactly like you've done physically, only logically. There's a MATRIX_HEIGHT and MATRIX_WIDTH defined in the header, but for things that aren't HUB75, they're not used super consistently. (The 3-panel audio spectrum analyzer is probably the most notable exception.) There is a combination of a macro (gfxbase.h:XY), a function (gxfbase.cc:XY) and a virtual method (gfxbase:xy) that may be in play. I THINK if you set MATRIX_HEIGHT and MATRIX_WIDTH to appropriate values for your frame, you should (might?) fall into the definition of xy() that handles the common xy-looping model you're describing. There's even ASCII art to describe it. So set those two #defines, confirm with your debugger that _height and _width are set correctly when GFXBase is constructed in gfxbase.cc:254 or so, sanity check that GetLEDCount() matches what you think it should be, and then see if our various matrix-aware code will play nice. We have a bunch of convenience operators like drawPixel() that takes x, y that should make manging them as a grid much more palatable. I've thought about adding virtual scrolling marquee text for hardware like the aforementioned wide panels of the 16x16 panels. I think we already have fonts in the project for Mesmerizer. Good luck!

…

On Sat, Jan 11, 2025 at 9:44 PM Gray Oron ***@***.***> wrote: Hi! I have a bunch of 4' long WS2812B strips mounted on strips of wood that are mounted with 3d printed clips, allowing me to use them in different configurations. I would love to use them as a matrix as well (to create wider effects or to display text). For the life of me I cannot figure out if there is an actual simple way to define a matrix zigzagged led strips in the projects (I have an din and dout on the ends, so I would just ideally chain them and inject some power ever so often). Help please! Oh, I am running them on Seeed Studio esp32-s3 which was not too hard to do with this project, but have not added too many effects so far... — Reply to this email directly, view it on GitHub <#676>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD37V5D4W7Q34NTSHQWD2KHQIFAVCNFSM6AAAAABVATD666VHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXHAYDSNZVHA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[GrayOrion]

Thank you @robertlipe!
That gives me plenty to work from, good to know there is no (clear) implementation of that yet.
I might just have to document my steps well and post something here for future reference, or better yet, suggest some code/config changes :-)

[robertlipe]

I think you'll find all the needed pieces there. The gate was on "simple". Certainly the spectrum project uses all the parts as the 8x8 panels that it uses are wired in the same layout as you're describing. It sets up width and height at https://github.com/PlummersSoftwareLLC/NightDriverStrip/blob/08d1895783da955584534930ee3e329b6b87ec2e/include/globals.h#L597 It is then able to address them at the software level like an x/y frame buffer. Good luck!

…

On Mon, Jan 13, 2025, 3:20 PM Gray Oron ***@***.***> wrote: Thank you @robertlipe <https://github.com/robertlipe>! That gives me plenty to work from, good to know there is no (clear) implementation of that yet. I might just have to document my steps well and post something here for future reference, or better yet, suggest some code/config changes :-) — Reply to this email directly, view it on GitHub <#676 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD3YP3IP2R2ESL2Y6R4T2KQUYHAVCNFSM6AAAAABVATD666VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCOBSGQ4DINI> . You are receiving this because you were mentioned.Message ID: <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/676/comments/11824845 @github.com>

[GrayOrion]

I am very confused about this error message while trying to run PatternSMFire2021 on a simple led matrix:

Compiling .pio\build\demo\src\effects.cpp.o
In file included from include/globals.h:1820,
                 from include/effectsupport.h:35,
                 from src/effects.cpp:31:
include/effects/matrix/PatternSMFire2021.h: In member function 'virtual void PatternSMFire2021::Draw()':
include/gfxbase.h:293:26: error: 'xy' was not declared in this scope
         #define XY(x, y) xy(x, y)
                          ^~
include/effects/matrix/PatternSMFire2021.h:63:34: note: in expansion of macro 'XY'
                 nblend(g()->leds[XY(x, MATRIX_HEIGHT - 1 - y)], GetBlackBodyHeatColor(Col/255.0f, g()->ColorFromCurrentPalette(0, Bri)).fadeToBlackBy(255-Bri), pcnt);
                                  ^~
include/gfxbase.h:293:26: note: suggested alternative: 'y'
         #define XY(x, y) xy(x, y)
                          ^~
include/effects/matrix/PatternSMFire2021.h:63:34: note: in expansion of macro 'XY'
                 nblend(g()->leds[XY(x, MATRIX_HEIGHT - 1 - y)], GetBlackBodyHeatColor(Col/255.0f, g()->ColorFromCurrentPalette(0, Bri)).fadeToBlackBy(255-Bri), pcnt);
                                  ^~
*** [.pio\build\demo\src\effects.cpp.o] Error 1
Retrieved `.pio\build\demo\lib645\ESP Async WebServer\AsyncWebSocket.cpp.o' from cache
================================================================================================================================================================ [FAILED] Took 32.01 seconds ================================================================================================================================================================
Environment    Status    Duration
-------------  --------  ------------
demo           FAILED    00:00:32.009

As the xy virtual method is defined right above the calling 'include' in gfxbase.h:

virtual uint16_t xy(uint16_t x, uint16_t y) const
    {
        if (x & 0x01)
        {
            // Odd rows run backwards
            uint8_t reverseY = (_height - 1) - y;
            return (x * _height) + reverseY;
        }
        else
        {
            // Even rows run forwards
            return (x * _height) + y;
        }
    }

    // This is an optimization that allows us to use direct math for the XY lookup when using the matrix, where
    // it's a very simple layout.  Others may need to override this function.  Using a #define here allows
    // us to avoid an extra virtual function call in the inner loop of the effects.

    #if USE_HUB75
        #define XY(x, y) ((y) * MATRIX_WIDTH + (x))
    #elif HELMET
        #define XY(x, y) xy(x, MATRIX_HEIGHT - 1 - y)           // Invert the Y axis for the helmet display
    #else
        #define XY(x, y) xy(x, y)
    #endif

I even added:

I am trying to run "EFFECT_MATRIX_SMFIRE2021" with a simple led matrix...
So many different places to make sure that everything needed is included, with vscode, copilot, etc unable to pinpoint where am I missing the plot:

    #elif LEDSTRIP

        ADD_EFFECT(EFFECT_STRIP_STATUS, StatusEffect, CRGB::White);
        ADD_EFFECT(EFFECT_MATRIX_SMFIRE2021,        PatternSMFire2021);

    #else
        ADD_EFFECT(EFFECT_MATRIX_SMFIRE2021,        PatternSMFire2021);
        // Animate a simple rainbow palette by using the palette effect on the built-in rainbow palette
        ADD_EFFECT(EFFECT_STRIP_RAINBOW_FILL, RainbowFillEffect, 6, 2);                    // Simple effect if not otherwise defined above
        ADD_EFFECT(EFFECT_STRIP_FIRE_FAN, FireFanEffect, RainbowColors_p, NUM_LEDS, 1, 10, 400, 2, NUM_LEDS - 20, Sequential, true, false);
        
        
    #endif

[rbergen]

@GrayOrion Do you have your code pushed to a branch in your fork that I can pull and look at?

I ask because I'm almost certain this is caused by something that's not included in the code you're now sharing. Phrased one way, there are 41 projects/environments that compile successfully with the current codebase, so somehow some change you made got things misaligned.

If I can comprehensively examine the state of what it is you try to compile then I can see if I can figure out exactly where that happened.

[GrayOrion]

Sorry for the slow response, work took over my life in the past few days...
My branch is at https://github.com/GrayOrion/NightDriverStrip/tree/graysMatrixTest
I added my environment changes as well (platformio.ini) so that you can see everything I might have screwed up...
It is most likely something I missed on the C++ side of things rather than anything with the actual project... :-(

Thanks for being willing to help a newb on a possibly really silly mistake or misunderstanding!

[robertlipe]

Without looking, Rutger, I'll bet he's trying to make that fire effect work on a hardware hardware configuration that approximately like your spectrum build. For you, I'll be a fireplace and the aspect ratio may act wacky at 48x16, but I'll bet that's the approximate goal. (I also get that doing it from scratch isn't YOUR project. 😊)

…

On Sun, Jan 19, 2025, 4:02 AM Rutger van Bergen ***@***.***> wrote: @GrayOrion <https://github.com/GrayOrion> Do you have your code pushed to a branch in your fork that I can pull and look at? I ask because I'm almost certain this is caused by something that's not included in the code you're now sharing. Phrased one way, there are 41 projects/environments that compile successfully with the current codebase, so somehow some change you made got things misaligned. If I can comprehensively examine the state of what it is you try to compile then I can see if I can figure out exactly where that happened. — Reply to this email directly, view it on GitHub <#676 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD3ZXU7CKXPZUZDQ2XAD2LNZ3ZAVCNFSM6AAAAABVATD666VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCOBYGA2TGNY> . You are receiving this because you were mentioned.Message ID: <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/676/comments/11880537 @github.com>

[rbergen]

I assume you're right about that, but without access to the code OP's now trying to compile, I can't see what "crossing wires" changes cause the build error they're seeing.

So, I'll go back to waiting for the OP's response, for now. 😊

[robertlipe]

Sure, it's always best to have complete code and a buildable reproduction step for this kind of thing. I had a few minutes and thought I'd look into it. I see only pretty small changes but the demo project still links. I don't have a matrix within reach to confirm working, but from your description, I expected a compile error. I have LOTS of experience with those, so was willing to help take a look. You changed the demo target, so that's what I built: $ pio run -e demo Compiles and links without issue Can you please provide steps to repro the compile failure so we can help you?

…

On Wed, Jan 22, 2025 at 1:14 AM Gray Oron ***@***.***> wrote: Sorry for the slow response, work took over my life in the past few days... My branch is at https://github.com/GrayOrion/NightDriverStrip/tree/graysMatrixTest I added my environment changes as well (platformio.ini) so that you can see everything I might have screwed up... It is most likely something I missed on the C++ side of things rather than anything with the actual project... :-( — Reply to this email directly, view it on GitHub <#676 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCSD3YUNJNZH5F3TMVUPP32L5AN5AVCNFSM6AAAAABVATD666VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCOJRGE4TCMA> . You are receiving this because you were mentioned.Message ID: <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/676/comments/11911910 @github.com>

[robertlipe]

OK, I found the problem. Following GitHub's instructions lost the branch you were working on. Got it. It now compiles without error. This time, instead of being in bed without strips, I'm in a dentist parking lot without strips, but I can explain the needed changes to get to a succesful build. Your approach of reusing DEMO kind of hits a few other branches in our configuration system. The lines between 'use 2d', "MESMERIZER", and USE_HUB75 are...fuzzy. globals.h and effects.* make it clear that you're the first to try this in a while. ledstripeffects.h and ledmatrixeffects.h are mutually exclusive. For better and worse, since you chose -DDEMO, you fell into the stripeffects.h path. This means that perfectly lovely definition of xy() in *matrixefffects.h was never seen because it was nuked via the USE_HUB75 at the top. (To split hairs, "provide 2d support" and "USE_HUB75" aren't quite the same thing, but here we are. If you're serious about pushing this upstream, we'll figure out how to solve some of these things better.) So the trick is to provide an xy() in ledstripeffects.h that provides the snaking effect you need. I just copied the one from *matrix*. It's looking for class members that don't exist, so I hardwired it to use the #defines you provided in globals.h. That made it blow up someewhere else, so I had to reset _width and _hegigtht when done. src/effects.cpp has similar logic that trips you up. There was a warning that you didn't report, but it provided hints probably only recognizable to those of us that have implemented effects. src/effects.cpp is the file that uses the effect name from p*ini to decide which effects/{strip,matrix}/file to include and when to add it to the needed arrays. You found a way to jam in the latter, via sticking an ADD_EFFECT, but you missed the #include. I just bypassed all that clever stuff and stuffed it in unconditionally, which is totally wrong if you were to try to upstream this, but it's fine for a personal project if you're building only one project from this tree...and if you're not, you have the hint yo need. You probably missed it, but you can put the ADD_EFFECT stuff in a personal custom_effects.h and bypass all that too-clever stuff for defines. That doesn't work for something you plan to upstream, but it's how I do my custom builds, and it works well. This does point out a hole in the scheme that there should also probably be a corresponding custom_effects_includes or something. Patch attached since I can't exactly run a p4 upload from here. I made it compile. Making it _work_ and making it awesome is back in your court for now. :-) If this is something you're interested in upstreaming, and I'd encourage you to, if we can get a nod from a reviewer on project name and loose agreement that we need to separate "1d" and "2d" along slightly different internal clues than we're using now, I'll help you add the needed goo to platform to create a new environment (you had the right idea reusing DEMO and that probably let you defer dodging several bullets...) I'm interested in landing this and will help whip all that into a shape to integrate it if you want to partner up on it. Use case: running many of our grand effects on WS281x panels and/or custom strip builds for that thing in the desert. RJL

…

On Wed, Jan 22, 2025 at 8:01 AM Robert Lipe ***@***.***> wrote: Sure, it's always best to have complete code and a buildable reproduction step for this kind of thing. I had a few minutes and thought I'd look into it. I see only pretty small changes but the demo project still links. I don't have a matrix within reach to confirm working, but from your description, I expected a compile error. I have LOTS of experience with those, so was willing to help take a look. You changed the demo target, so that's what I built: $ pio run -e demo Compiles and links without issue Can you please provide steps to repro the compile failure so we can help you? On Wed, Jan 22, 2025 at 1:14 AM Gray Oron ***@***.***> wrote: > Sorry for the slow response, work took over my life in the past few > days... > My branch is at > https://github.com/GrayOrion/NightDriverStrip/tree/graysMatrixTest > I added my environment changes as well (platformio.ini) so that you can > see everything I might have screwed up... > It is most likely something I missed on the C++ side of things rather > than anything with the actual project... :-( > > — > Reply to this email directly, view it on GitHub > <#676 (reply in thread)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ACCSD3YUNJNZH5F3TMVUPP32L5AN5AVCNFSM6AAAAABVATD666VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCOJRGE4TCMA> > . > You are receiving this because you were mentioned.Message ID: > <PlummersSoftwareLLC/NightDriverStrip/repo-discussions/676/comments/11911910 > @github.com> >

[rbergen]

Many thanks @robertlipe for not only looking into this but also providing a solution and giving pointers on paths forward before I even had a chance to properly look at this.

Unfortunately, the patch you attached got lost in GitHub's email handling magic. You can send it to me via "regular" email, after which I will press the correct GitHub buttons to attach it to this discussion in a way that works in the GitHub realm.

[robertlipe]

You're welcome. It may be clear that this is relevant to my interests and something I'd already investigated for other projects. I want to do things like provide a custom x-y that builds a radial map so we can display some of our effects on the circular panels or the brain-bleeding math I've been tinkering with to make Hexagons awesome and other things that we'd like to treat as an X/Y addressable thing instead of a strip.

I was going to just stick this as a PR to GrayOrion's tree, but 'gh pr new' looked like it was going to send it to the Plummer tree, which has happened before and takes us both effort to unwind.

There's certainly no magic to the patch itself. It's blunt force just to show the work that needs to be done and NOT at all a model of how to do it "right"... if we decide that doing it right is a thing we want to tackle.

This is basically what I'd do to get all that nasy stuff out of the way to actually focus on the effects and then loop back to this kind of glue pennance.

[GrayOrion]

Thank you, @robertlipe! Amazing help and guidance, above and beyond!
That is amazing, and yes, I am happy to work on it to make it usable to others, with guidance on the framework side of things.
I know about the costum_effects.h but I was trying to do as little as possible to start with...
Also, missing in my branch - I brute forced including PatternSMFire2021.h at the top of effects.cpp, and then got the xy error
But that is beside the point now. I will follow through your guidance a bit more closely this weekend and suggest some proper changes / ask for guidance and hopefully we can get an awesome zigzag matrix framework.
FYI, one of the reasons I want to use Nightdriver is for our local burner event where I will hopefully run the leds on the walls of a geodesic dome... So interested in 2d, but also 3d in the future (though I am only lighting the "wrapper" this year, so I can treat it as a 2d surface, I guess...

[GrayOrion]

You're welcome. It may be clear that this is relevant to my interests and something I'd already investigated for other projects. I want to do things like provide a custom x-y that builds a radial map so we can display some of our effects on the circular panels or the brain-bleeding math I've been tinkering with to make Hexagons awesome and other things that we'd like to treat as an X/Y addressable thing instead of a strip.

I was going to just stick this as a PR to GrayOrion's tree, but 'gh pr new' looked like it was going to send it to the Plummer tree, which has happened before and takes us both effort to unwind.

There's certainly no magic to the patch itself. It's blunt force just to show the work that needs to be done and NOT at all a model of how to do it "right"... if we decide that doing it right is a thing we want to tackle.

This is basically what I'd do to get all that nasy stuff out of the way to actually focus on the effects and then loop back to this kind of glue pennance.

Robert, how do I get in touch with you?
I am afraid your profile is invisible to me...
I would love it if you connect with me through the email in my profile, and we can plot the general steps and path to make matrices work for led strips, as well as other oddly shaped skins... :-)

[robertlipe]

diff --git a/include/ledstripeffect.h b/include/ledstripeffect.h
index eb18978f..bf196dcd 100644
--- a/include/ledstripeffect.h
+++ b/include/ledstripeffect.h
@@ -637,5 +637,19 @@ class LEDStripEffect : public IJSONSerializable
 
         return false;
     }
+#define _width MATRIX_WIDTH
+#define _height MATRIX_HEIGHT
+    uint16_t xy(uint16_t x, uint16_t y) const // override
+    {
+        // Note the x,y are unsigned so can't be less than zero
+        if (x < _width && y < _height)
+            return y * MATRIX_WIDTH + x;
+
+        debugE("%s", str_sprintf("Invalid index in xy: x=%d, y=%d, NUM_LEDS=%d", x, y, NUM_LEDS).c_str());
+        return 0;
+    }
+#undef _width
+#undef _height
+
 };
 
diff --git a/src/effects.cpp b/src/effects.cpp
index e8d827cc..19d41583 100644
--- a/src/effects.cpp
+++ b/src/effects.cpp
@@ -57,6 +57,7 @@
 // Externals
 //
 
+    #include "effects/matrix/PatternSMFire2021.h"
 #if USE_HUB75
     #include "ledmatrixgfx.h"
 

g.patch

[GrayOrion]

So simple... thanks @robertlipe
OK, got this working, obviously not very complicated to brute force it.
And thanks to your explanation I understand a bit better the different "flavors" and how they flow (and how the config build flows)...
I will have to take a look at the code style / preferences for this project a bit more, not sure how peeps feel about different types of polymorphism and inheritance, and since I spent most of my career in Java (don't hate me!), I don't have a strong opinions or the knowledge to argue for one when it comes to C++. I will follow others suggestions! In any case, I think it would be best we take this off this thread to DM, and then link here a branch once we start working on it (which might be a couple of weeks for me as I will be away on vacation next week).

[rbergen]

Glad to hear you got this working, and I also appreciate your intention to contribute to the main codebase - stating immediately that there are no hard feelings if that doesn't end up happening.

As one of the peeps on this project, a few notes on coding styles etc.:

• The codebase shows that this project has existed for 7-ish years, in the sense that the coding style has developed over time, but not been retroactively applied to existing code for the sake of it.
• More recent additions lean more strongly towards "modern C++" (we're currently compiling C++20), certainly driven in part by @robertlipe's suggestions in that direction.
• That said, we try not to be too anal about coding style as such, certainly when it concerns new code (as in: new source code files.) Again, if you leave opportunities to use modern C++ unused, do expect @robertlipe to point that out in his review of your code. :)
• Changes to existing code should be as small as possible to implement the (functional) change you're aiming for, and follow the coding style of that specific file.
• Bigger overhauls (refactors) of code are accepted provided they serve the purpose, but you do then "own" the change, meaning you need to make sure the code you touch still works and in principle not just for the one project you use. (CI does help with that by compiling all 40+ configurations when you push or open a PR.)

Concerning Java: a university professor I am still grateful to replaced an antiquated part of the Computer Science curriculum (we were going to learn about dial-up modeming into BBSs using Telix in 1996) with Java, correctly predicting it would become one of the languages of the Internet age. I also made my first "developer job" guilders (Dutch currency before the Euro) using it, and I fully acknowledge that my favorite Internet age language C# was very, very heavily inspired by it.
So, you don't have to expect any looking down on Java from me at least. :)